1/*
2 * Copyright (c) 2006 Paul Richards <paul.richards@gmail.com>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21/*!
22 * \file libtheoraenc.c
23 * \brief Theora encoder using libtheora.
24 * \author Paul Richards <paul.richards@gmail.com>
25 *
26 * A lot of this is copy / paste from other output codecs in
27 * libavcodec or pure guesswork (or both).
28 *
29 * I have used t_ prefixes on variables which are libtheora types
30 * and o_ prefixes on variables which are libogg types.
31 */
32
33/* FFmpeg includes */
34#include "libavutil/intreadwrite.h"
35#include "libavutil/log.h"
36#include "avcodec.h"
37
38/* libtheora includes */
39#include <theora/theora.h>
40
41typedef struct TheoraContext{
42    theora_state t_state;
43} TheoraContext;
44
45/*!
46    Concatenates an ogg_packet into the extradata.
47*/
48static int concatenate_packet(unsigned int* offset, AVCodecContext* avc_context, const ogg_packet* packet)
49{
50    char* message = NULL;
51    uint8_t* newdata = NULL;
52    int newsize = avc_context->extradata_size + 2 + packet->bytes;
53
54    if (packet->bytes < 0) {
55        message = "ogg_packet has negative size";
56    } else if (packet->bytes > 0xffff) {
57        message = "ogg_packet is larger than 65535 bytes";
58    } else if (newsize < avc_context->extradata_size) {
59        message = "extradata_size would overflow";
60    } else {
61        newdata = av_realloc(avc_context->extradata, newsize);
62        if (newdata == NULL) {
63            message = "av_realloc failed";
64        }
65    }
66    if (message != NULL) {
67        av_log(avc_context, AV_LOG_ERROR, "concatenate_packet failed: %s\n", message);
68        return -1;
69    }
70
71    avc_context->extradata = newdata;
72    avc_context->extradata_size = newsize;
73    AV_WB16(avc_context->extradata + (*offset), packet->bytes);
74    *offset += 2;
75    memcpy( avc_context->extradata + (*offset), packet->packet, packet->bytes );
76    (*offset) += packet->bytes;
77    return 0;
78}
79
80static av_cold int encode_init(AVCodecContext* avc_context)
81{
82    theora_info t_info;
83    theora_comment t_comment;
84    ogg_packet o_packet;
85    unsigned int offset;
86    TheoraContext *h = avc_context->priv_data;
87
88    /* Set up the theora_info struct */
89    theora_info_init( &t_info );
90    t_info.width = avc_context->width;
91    t_info.height = avc_context->height;
92    t_info.frame_width = avc_context->width;
93    t_info.frame_height = avc_context->height;
94    t_info.offset_x = 0;
95    t_info.offset_y = 0;
96    /* Swap numerator and denominator as time_base in AVCodecContext gives the
97     * time period between frames, but theora_info needs the framerate.  */
98    t_info.fps_numerator = avc_context->time_base.den;
99    t_info.fps_denominator = avc_context->time_base.num;
100    if (avc_context->sample_aspect_ratio.num != 0) {
101        t_info.aspect_numerator = avc_context->sample_aspect_ratio.num;
102        t_info.aspect_denominator = avc_context->sample_aspect_ratio.den;
103    } else {
104        t_info.aspect_numerator = 1;
105        t_info.aspect_denominator = 1;
106    }
107    t_info.colorspace = OC_CS_UNSPECIFIED;
108    t_info.pixelformat = OC_PF_420;
109    t_info.target_bitrate = avc_context->bit_rate;
110    t_info.keyframe_frequency = avc_context->gop_size;
111    t_info.keyframe_frequency_force = avc_context->gop_size;
112    t_info.keyframe_mindistance = avc_context->keyint_min;
113    t_info.quality = 0;
114
115    t_info.quick_p = 1;
116    t_info.dropframes_p = 0;
117    t_info.keyframe_auto_p = 1;
118    t_info.keyframe_data_target_bitrate = t_info.target_bitrate * 1.5;
119    t_info.keyframe_auto_threshold = 80;
120    t_info.noise_sensitivity = 1;
121    t_info.sharpness = 0;
122
123    /* Now initialise libtheora */
124    if (theora_encode_init( &(h->t_state), &t_info ) != 0) {
125        av_log(avc_context, AV_LOG_ERROR, "theora_encode_init failed\n");
126        return -1;
127    }
128
129    /* Clear up theora_info struct */
130    theora_info_clear( &t_info );
131
132    /*
133        Output first header packet consisting of theora
134        header, comment, and tables.
135
136        Each one is prefixed with a 16bit size, then they
137        are concatenated together into ffmpeg's extradata.
138    */
139    offset = 0;
140
141    /* Header */
142    theora_encode_header( &(h->t_state), &o_packet );
143    if (concatenate_packet( &offset, avc_context, &o_packet ) != 0) {
144        return -1;
145    }
146
147    /* Comment */
148    theora_comment_init( &t_comment );
149    theora_encode_comment( &t_comment, &o_packet );
150    if (concatenate_packet( &offset, avc_context, &o_packet ) != 0) {
151        return -1;
152    }
153
154    /* Tables */
155    theora_encode_tables( &(h->t_state), &o_packet );
156    if (concatenate_packet( &offset, avc_context, &o_packet ) != 0) {
157        return -1;
158    }
159
160    /* Clear up theora_comment struct */
161    theora_comment_clear( &t_comment );
162
163    /* Set up the output AVFrame */
164    avc_context->coded_frame= avcodec_alloc_frame();
165
166    return 0;
167}
168
169static int encode_frame(
170    AVCodecContext* avc_context,
171    uint8_t *outbuf,
172    int buf_size,
173    void *data)
174{
175    yuv_buffer t_yuv_buffer;
176    TheoraContext *h = avc_context->priv_data;
177    AVFrame *frame = data;
178    ogg_packet o_packet;
179    int result;
180
181    assert(avc_context->pix_fmt == PIX_FMT_YUV420P);
182
183    /* Copy planes to the theora yuv_buffer */
184    if (frame->linesize[1] != frame->linesize[2]) {
185        av_log(avc_context, AV_LOG_ERROR, "U and V stride differ\n");
186        return -1;
187    }
188
189    t_yuv_buffer.y_width = avc_context->width;
190    t_yuv_buffer.y_height = avc_context->height;
191    t_yuv_buffer.y_stride = frame->linesize[0];
192    t_yuv_buffer.uv_width = t_yuv_buffer.y_width / 2;
193    t_yuv_buffer.uv_height = t_yuv_buffer.y_height / 2;
194    t_yuv_buffer.uv_stride = frame->linesize[1];
195
196    t_yuv_buffer.y = frame->data[0];
197    t_yuv_buffer.u = frame->data[1];
198    t_yuv_buffer.v = frame->data[2];
199
200    /* Now call into theora_encode_YUVin */
201    result = theora_encode_YUVin( &(h->t_state), &t_yuv_buffer );
202    if (result != 0) {
203        const char* message;
204        switch (result) {
205            case -1:
206                message = "differing frame sizes";
207                break;
208            case OC_EINVAL:
209                message = "encoder is not ready or is finished";
210                break;
211            default:
212                message = "unknown reason";
213                break;
214        }
215        av_log(avc_context, AV_LOG_ERROR, "theora_encode_YUVin failed (%s) [%d]\n", message, result);
216        return -1;
217    }
218
219    /* Pick up returned ogg_packet */
220    result = theora_encode_packetout( &(h->t_state), 0, &o_packet );
221    switch (result) {
222        case 0:
223            /* No packet is ready */
224            return 0;
225        case 1:
226            /* Success, we have a packet */
227            break;
228        default:
229            av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed [%d]\n", result);
230            return -1;
231    }
232
233    /* Copy ogg_packet content out to buffer */
234    if (buf_size < o_packet.bytes) {
235        av_log(avc_context, AV_LOG_ERROR, "encoded frame too large\n");
236        return -1;
237    }
238    memcpy(outbuf, o_packet.packet, o_packet.bytes);
239
240    return o_packet.bytes;
241}
242
243static av_cold int encode_close(AVCodecContext* avc_context)
244{
245    ogg_packet o_packet;
246    TheoraContext *h = avc_context->priv_data;
247    int result;
248    const char* message;
249
250    result = theora_encode_packetout( &(h->t_state), 1, &o_packet );
251    theora_clear( &(h->t_state) );
252    switch (result) {
253        case 0:/* No packet is ready */
254        case -1:/* Encoding finished */
255            return 0;
256        case 1:
257            /* We have a packet */
258            message = "gave us a packet";
259            break;
260        default:
261            message = "unknown reason";
262            break;
263    }
264    av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed (%s) [%d]\n", message, result);
265    return -1;
266}
267
268static const enum PixelFormat supported_pixel_formats[] = { PIX_FMT_YUV420P, PIX_FMT_NONE };
269
270/*! AVCodec struct exposed to libavcodec */
271AVCodec libtheora_encoder =
272{
273    .name = "libtheora",
274    .type = CODEC_TYPE_VIDEO,
275    .id = CODEC_ID_THEORA,
276    .priv_data_size = sizeof(TheoraContext),
277    .init = encode_init,
278    .close = encode_close,
279    .encode = encode_frame,
280    .pix_fmts = supported_pixel_formats,
281    .long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"),
282};
283